|Title||»||Before starting a new Import > Results we need to check that Decoys are representative of the Targets|
I have just been looking at a paper where the mProphet model was totally messed up because the user generated decoys and then edited the file enough that the decoys were easily differentiated from the Targets just by transition count (6 in targets and 10 in decoys) which lead to "Coelution count" scores that became the dominant distinguishing features.
I am personally always very careful to regenerate my decoys once I feel the set of Targets is complete and just before I start importing data I intend to build an mProphet model for. But, though, this may be "best practice" we don't provide much guidance in the software to ensure others are doing this.
At File > Import > Results time we should check the decoys and decide whether they still represent the targets and offer to regenerate them if they do not, while also checking to ensure there is not already data imported with the existing decoys.
Not exactly sure how best to make this check, but it needs to be done to protect users from what I have recently seen as a serious and undetected error in data processing that nearly got published.